The machine learning model for detecting foreign objects was tested by the team.
The research team from the Vietnam Aviation Academy has outlined a realistic airport simulation including the entire terminal, aircraft, runway, telescopic tube, lighting system (day and night simulation). Besides, the team arranged cameras to detect objects even along the runway.
Testing the system on a model airport is very different from the real airport because of the distance from the camera position (satisfying safety conditions) to the object (edge length over 3 cm) on the runway is very large, sometimes up to hundreds of meters. Therefore, the camera system needs higher resolution to identify objects and needs a computer system with faster data processing speed. In Vietnam, currently airports do not use automatic systems to detect foreign objects, most use manual methods (airports mobilize people to control and collect foreign objects in designated runways, taxiways, aprons areas).
The research team tested a machine learning model with images in well-lit conditions, resulting in over 99% accuracy in detecting foreign objects. As for images with noise, that is, in low light conditions, dust, rain and wind and etc. the model operates with lower accuracy, about 70-80% on average. Currently, the research team's products are designed to detect objects on the ground. In the coming time, the group will continue to research and develop similar functions for airborne objects, soon test production and apply at domestic airports. |